**2.2 Forest inventory**

*Protected Areas, National Parks and Sustainable Future*

*Description of the 16 sites studied at La Mauricie National Park of Canada.*

low to ensure eastern white pine renewal, thus preventing the park from reaching its objective of maintaining or restoring ecological integrity [40]. Burning prescriptions were defined using the Canadian Forest Fire Danger Rating System [41] and the software FBP97 for forecasting fire behaviour [42]. Prescribed burnings were carried out during spring because burning conditions are more suitable before bud flushing of broadleaved trees and shrubs [43]. When conditions were appropriate, fire was ignited using burners (driptorch; **Figure 2**) or a helicopter equipped with a Premo MK3 aerial ignition device. Low-intensity surface fires were isolated and controlled with natural and artificial firebreaks. Flame height and length were recorded during each prescribed burning event by the park's staff and were used to estimate fire intensity based on the Canadian forest fire behaviour prediction system [41] (**Table 1**). For low-intensity surface fires, these classes range from 1 (frontal fire intensity < 10 kW/m; flame length < 0.2 m; flame height < 0.1 m) to 5 (frontal fire intensity > 4000 kW/m; flame length >3.5 m; flame height >2.5 m). In our study, fire intensity in burned sites mostly belongs to class

*Parks Canada crew using a driptorch to run a prescribed burning experiment in a white pine stand at La* 

**42**

**Figure 2.**

**Table 1.**

*Mauricie National Park of Canada.*

Three 400-m2 circular plots located 50 m apart along a transect and at a minimum distance of 50 m from stand or treatment edges were set up in each stand to describe the forest environment. In each plot, we recorded the slope (%), altitude (m), surface deposit, drainage, and thickness of the soil organic layer (litter and humus) (**Table 1**).

Species, diameter at breast height (hereafter DBH), and decay class of each standing tree or snag ≥9.1 cm at DBH were recorded. Decay classes were determined according to Hunter classification [44], which recognizes nine classes for trees (1: alive and 2: declining) and snags (3: dead tree with bark intact up to 9: stump). Because most pines were large and tall, their density was rather low and, to get more accurate estimates of their basal area, we enlarged the sampled plots up to 1200 m<sup>2</sup> (radius = 19.55 m). In each 400-m2 plot, four smaller plots of 25 m2 (radius = 2.82 m) and four micro plots of 4 m2 (radius = 1.13 m) were established at 8.46 m from the plot centre, in each cardinal direction. Saplings and seedlings were recorded in the 25 and 4-m2 plots, respectively. Saplings were defined as young trees in which DBH ranged between 1 and 9 cm, whereas seedlings were very young trees with DBH smaller than 1 cm [45]. Sapling DBH was measured and seedling height was recorded into 5-cm classes. Eastern white pine relative dominance was estimated on the basis of its relative basal area (hereafter BA, in m2 /ha) in 1200-m<sup>2</sup> plots, in relation to BA of other tree species estimated in the 400-m2 plots.

## **2.3 Statistical analysis**

As stands had not been sampled before treatment, the short-term effects of prescribed burning were assessed using the percentage of recent tree or sapling mortality in 1- to 7-year-old burns (older burns could not represent short-term effects of prescribed burning) and compared to unburned stands. Tree BA and sapling density (stems/ha) were calculated for eastern white pine, balsam fir, spruces, and broadleaved species. Then, the percentages of recent mortality (Hunter classes 3 and 4) were calculated for both burned and unburned stands. Student's *t*-tests were used to compare recent mortality of trees and saplings in both stand types. We also used Student's *t*-tests to compare seedling density in burned and unburned stands. Sites burned in 2004 and 2005 were excluded from the seedling analysis because no seed crop had occurred after the treatment, thus precluding the establishment of regeneration in these stands. Logarithmic transformations (log x + 1) were used to normalize the distributions and stabilize variances when necessary. When transformations did not achieve equality of variances, we used results obtained with Satterthwaite's approximate *t*-test, a method that belongs to the Behrens-Welch family [46]. Analyses were performed using SAS software v. 9.1. [47].
